Method and system for quantifying cardiopulmonary-emotional interaction mechanism based on symbolic transfer entropy

Abstract

The application discloses a heart-lung-emotion interaction mechanism quantification method and system based on symbolic transfer entropy, and aims to overcome the problem of low time resolution of traditional analysis methods by generating continuous dynamic synchronous curves DSC, so that the dynamic characteristics of heart-lung interaction can be captured and the causal relationship between the dynamic characteristics and emotional states can be analyzed. The method first collects electrocardio and respiration signals, generates high time resolution heart-lung dynamic synchronous curves through multi-dimensional parameter space fusion, calculates symbolic transfer entropy in combination with emotional annotation sequences, and quantifies the bidirectional information flow direction and intensity between the heart-lung system and emotional dimensions. The event-related analysis method is used to analyze the dynamic interaction mode of "drive-response", and to identify whether the current psychological-physiological state is driven by physiological rhythm or regulated by cognitive emotion. The application can not only capture transient heart-lung synchronous events, but also accurately determine whether the current state is driven by heart-lung synchronization to experience emotion or regulated by cognitive emotion to physiological rhythm.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of medicine and signal processing technology, specifically to a method and system for quantifying the cardiopulmonary-emotion interaction mechanism based on symbolic transfer entropy. Background Technology

[0002] The interaction between the cardiovascular and respiratory systems is a fundamental physiological process in the human body and is finely regulated by the autonomic nervous system (ANS). In addition to the well-known frequency modulation phenomenon of respiratory sinus arrhythmia (RSA), studies based on nonlinear dynamics have shown that the cardiopulmonary system can be regarded as two weakly coupled self-sustaining chaotic oscillators [1]. This perspective makes phase synchronization an important window for studying the interaction between the two systems, that is, the phase of the two systems is locked at a specific point. n:m On an integer ratio, the magnitude may remain chaotic and uncorrelated.

[0003] To visualize this synchronization phenomenon, Schäfer et al. proposed the cardiopulmonary synchronization chart (CRS), which uses the straight lines formed by the relative phases in the chart to show the stability of the phase, and has become a classic tool in this field [2]. However, transforming the qualitative visualization tool CRS into a reliable quantitative indicator still faces many challenges in the existing technical solutions. The current mainstream quantitative methods mainly include phase regression analysis and histogram quantification [3, 5-6]. Phase regression determines synchronization by detecting whether the phase difference within a specific window remains within a threshold. Although it has a good time resolution, its results are highly dependent on the user-defined threshold and window length, and it is difficult to distinguish between true phase lock and "pseudo-synchronization" caused by close frequency. Histogram quantification detects synchronization by analyzing the statistical characteristics of the phase distribution, which usually requires a long analysis window (e.g., more than 20 respiratory cycles), resulting in low time resolution and inability to capture short-lived, transient synchronization events. In addition, most existing methods use fixed parameters for analysis, which is difficult to adapt to the non-stationarity and adaptability of physiological signals. For example, during a continuous recording, the cardiopulmonary system may switch between different synchronization ratios (such as 3:1, 5:2, etc.) or the synchronization strength may change dynamically. Traditional methods using a single fixed window or threshold often only provide binary (synchronized / asynchronous) classification results and lack a continuous, dynamic measure of coupling strength.

[0004] However, existing cardiopulmonary coupling analysis methods have significant limitations in practical applications such as intelligent emotion monitoring, biofeedback therapy, and human-computer interaction safety. Current mainstream technologies primarily focus on "correlation" detection, only determining whether cardiopulmonary synchronization occurs concurrently with a certain emotional state, but failing to analyze the causal driving relationship between the two. For example, in closed-loop psychological intervention and stress regulation systems, this lack of directional criteria prevents the system from determining whether the current high arousal state is driven by emotional fluctuations due to changes in cardiopulmonary synchronization patterns (i.e., a "bottom-up" pattern) or by autonomic nervous system responses triggered by changes in brain cognition (i.e., a "top-down" pattern). Lacking this crucial decision-making basis, the system cannot intelligently choose between "breathing induction" or "cognitive guidance" for precise regulation, limiting the effectiveness of intervention strategies. Transfer entropy (TE), as a nonparametric statistic based on information theory, can quantify the directed information flow between two processes without any model assumptions, making it highly suitable for analyzing nonlinear coupled systems. In particular, Symbolic Transfer Entropy (STE) not only inherits the advantages of TE by transforming continuous time series into discrete symbolic sequences, but also greatly enhances the robustness to outliers and measurement noise, while being more computationally efficient

[11] .

[0005] Furthermore, traditional methods also face accuracy limitations in identifying high-risk states (such as driver fatigue monitoring, high-pressure operation monitoring, and game interaction). For example, emotions like fear (negative) and ecstasy (positive) are both accompanied by high arousal, and simple synchronization intensity indicators often fail to effectively distinguish between these two distinct psychological states. Existing technologies lack sophisticated analytical methods for the dynamic evolution of cardiopulmonary synchronization (e.g., whether synchronization levels increase with emotional arousal or decrease with emotional valence), making it impossible to utilize the inherent differences in cardiopulmonary interaction mechanisms to eliminate false positive alarms in monitoring (such as misreporting extreme excitement as panic), thus limiting the application value of physiological signals in complex scenarios.

[0006] Therefore, there is an urgent need for a comprehensive analytical method that can quantify the dynamic intensity of cardiopulmonary synchronization, information flow, and analyze response patterns. This method not only needs high temporal resolution to capture transient synchronization events, but also needs to be able to analyze the dominant direction and specific mechanisms of "heart-brain interaction," thereby providing key quantitative indicators and decision-making basis for distinguishing complex emotion categories, guiding personalized biofeedback training, and constructing high-precision emotional safety monitoring systems. Summary of the Invention

[0007] In view of this, the present invention provides a method and system for quantifying the cardiopulmonary-emotion interaction mechanism based on symbolic transfer entropy, which can quantify the information flow between cardiopulmonary synchronization and emotional experience. Combining the high temporal resolution of DSC and the robustness of STE to noise and nonlinearity, it is suitable for analyzing complex physiological-psychological signals and is easy to software-encode.

[0008] To achieve the above objectives, this invention provides a method for quantifying the cardiopulmonary-emotion interaction mechanism based on symbolic transfer entropy, the technical solution of which includes the following steps:

[0009] S1: While watching emotion-evoked videos, the subjects' ECG and respiratory signals were collected using wearable devices, and their subjective emotional valence and arousal changes were recorded using a continuous annotation tool based on a two-dimensional ring model to obtain an emotion annotation sequence.

[0010] S2: Combining ECG and respiratory signals, dynamically quantify the changes in cardiopulmonary synchronization over a period of time and generate a result curve, producing a dynamic cardiopulmonary synchronization curve (DSC). The DSC is based on the R peak time. Non-uniform time series.

[0011] S3: Resample the DSC and sentiment-labeled sequences to a frequency consistent with the sentiment sampling rate to obtain a time-synchronized standardized sequence. and .

[0012] S4: Calculation and The directed information flow between them, namely the symbolic transfer entropy (STE) value; S5: Use alternative data method to perform nonparametric statistical tests to establish the significance of causal connections. By comparing the bidirectional information flow characteristics, establish the causal prediction relationship between signals, thereby determining the dominant driving direction and identifying the driving signal and response signal.

[0013] S6: Based on the determined dominant driving direction, identify the moment when a significant change occurs in the driving signal to define a mutation event, and obtain the response waveform of the corresponding driving event in the response signal. Based on the changing trend of the response waveform relative to the driving event, construct a state classification criterion to distinguish complex emotions with different inducing mechanisms.

[0014] Furthermore, while subjects watched emotionally evoked videos, their ECG and respiratory signals were collected using wearable devices, and then processed as follows: S1-1: For ECG signals, a 32nd-order FIR filter is used to filter and remove baseline drift. Then, the R-peak time points are extracted using the QRS wave detection algorithm and processed into a point process sequence. S1-2: For the respiratory signal, firstly remove baseline drift and high-frequency noise, then use a bandpass filter to process the signal and extract the frequency band where the respiratory signal is located, and then use Hilbert transform to calculate its analytic signal, thereby obtaining the continuous instantaneous respiratory phase.

[0015] Furthermore, by combining ECG and respiratory signals, the changes in cardiopulmonary synchronization over a period of time are dynamically quantified and a result curve is generated, resulting in a dynamic cardiopulmonary synchronization curve (DSC). The specific steps are as follows: S201: Screening candidate synchronization ratio; S202: Construct a multidimensional dynamic quantitative analysis framework, including three dimensions: (1) cardiopulmonary synchronization ratio (Ratio), which iterates through all candidate synchronization ratios; (2) sensitivity threshold, which sets a sensitivity threshold δ with a value range of [1, 20] and a step size of 1; and (3) scale, which sets different observation window widths. L ; S203: Combine ECG and respiratory signals to construct a cardiopulmonary synchronization map and perform multi-scale synchronization quantization to obtain a cardiopulmonary synchronization curve; S204: For different values ​​of cardiopulmonary synchronization ratio and sensitivity threshold, the above algorithm is traversed to iteratively update the cardiopulmonary synchronization curve and obtain the dynamic cardiopulmonary synchronization curve DSC. The obtained dynamic cardiopulmonary synchronization curve DSC is a continuously changing time series, which is used to intuitively represent the strength and stability of cardiopulmonary phase coupling.

[0016] Further, S201: Screening candidate synchronization ratios, the specific steps are as follows: First, for each subject, the ratio of the mean heart rate to the mean respiratory rate throughout the entire recording was calculated, and then rounded to the nearest integer to obtain the initial synchronization ratio. Based on this, select the integer ratios of its left and right adjacent values ​​( and but That is, the baseline candidate set for one breath includes ( , , ; m is the heart rate parameter in the synchronization ratio, and n is the respiratory parameter in the synchronization ratio; m The value of is iterated from 1 to 3, for each . m The value, its corresponding n The range of values ​​for is determined by the boundary expansion of the benchmark candidate set: the lower bound is . The upper boundary is Select the one that falls within [ , All integers within the numerical range k , construct the current mAll candidate synchronization ratios under the value k : m .

[0017] Furthermore, in S203, the specific steps are as follows: Combining the ECG and respiratory signals obtained in step 1, a cardiopulmonary synchronization map is constructed. First, the time corresponding to each R peak is calculated. t k Relative phase sequence relative to the respiratory cycle F r : (1) in, f resp This is the respiratory phase. m This is the synchronization ratio parameter. This process converts two physiological signals from different modalities into a unified phase space, laying the foundation for subsequent synchronization analysis; relative phase sequence F r ( t k Defined in formula (1), the sequence length is N ,Right now k = 1, 2, …, N Regarding the time scale L (Values ​​are 2, 3, 4, 5), in the synchronization ratio n : m Below, define a sliding window. W i = { F r ( t i ), F r ( t i+1 ), …, F r ( t i+n L-1 )}, i = 1, 2, …, N - n L +1, calculate the maximum deviation of the relative phase within each window. This deviation reflects the stability of phase locking. The specific process is as follows: (1) Grouping data within a window, defining "rows": in Under synchronization ratio, sliding window W i Contains A series of consecutive relative phase points, which should be distributed in n On different phase branches, first, the data points within the window are sorted by index modulus. n Group it and split it into n Subgroups, i.e. n Lines, each line L The number of points represents the window width. L Each row of data set Defined as: , j = 0, 1, …, n -1(2) in F r ( t k ) is the relative phase value within the window.

[0018] (2) Calculate the maximum phase difference within a "row": For any two time points within a "row" k a and k b , k a , k b {1, 2,…, L},and k a ≠ k b Their relative phases are respectively F r ( t ka )and F r ( t kb Define the absolute difference between the two and the relationship between this difference and the period. m The smaller of the complements is the phase difference between the two points, that is: (3) in, k a , k b {1, 2, …, L},and k a ≠ k b , For Euclidean distance, To eliminate the jump interference caused by phase periodicity, the complementary distance across the periodic boundary is calculated. The phase difference between every two points within a "row" is then taken as the maximum phase difference for that "row," denoted as . : , k a , k b {1, 2, …, L}, k a ≠ k b (4) (3) Calculate the maximum value between "rows": After the maximum phase difference of each row in the sliding window is calculated, the maximum phase difference of the window is defined as the maximum phase difference of the "row". , j = 0, 1, …, n The maximum value of -1: (5) in, m , n Synchronization ratio parameter L The dimension is the width of the window, and each sliding window... W i Inside n "Line", each "line" has L One element; (4) Determining the synchronization state: Represents the current window W i The worst-case phase-locked condition is defined if and only if the value is less than the sensitivity threshold. Only when the tolerance limit is determined will a judgment be made. W i In a synchronized state, an iterative update mechanism is introduced, abandoning the traditional single fixed threshold selection, setting a sensitivity threshold δ with a value from 1 to 20 and a step size of 1; defining a sequence. S L In scale L The following is a synchronized time series with a length of N For each sliding window W i If its maximum phase difference D( W i , L , n , m The tolerance is less than the current sensitivity threshold. m / ( n · d If the window is not found, then the decision window will be determined.W i At the threshold level d If we select "Synchronize", then we can synchronize the time series. S L Zhongyu W i The corresponding time point S L ( t i ), S L ( t i+1 ), …, S L ( t i+n L-1 Then assign the value d If the maximum phase difference D within the window ( W i , L , n , m Not less than the tolerance corresponding to the current threshold m / ( n · d ), then S L ( t i ), S L ( t i+1 ),…, S L ( t i+n L-1 The value is assigned to 0; Synchronous Time Series S L Refresh mechanism: When the sliding window is refreshed by... W i Become W i+1 When, if D( W i+1 , L , n , m Less than m / ( n · d If ), then record S L ( t i+1 ), S L ( ti+2 ), …, S L ( t i+n L )for d If D( W i+1 , L , n , m Not less than m / ( n · d If ), then retain. S L ( t i+1 ), …, S L ( t i+n L-1 The value of ) remains unchanged, and the record is maintained. S L ( t i+n L The value is 0; that is, in the synchronized time series S L The maximum synchronization threshold is retained during the sliding process; when taking i For 1, 2, …, N - n L When +1 is added, a synchronization time series reflecting the synchronization level can be obtained. S L ; (5) Multi-scale fusion: different time scales L ,right S L The sequences are weighted, and a weight vector is defined. Q =[ oh L1 , oh L2 , …, oh Ll ] l For scale L The number of possible values, and S L correspond, oh Li The value of varies with i The window width increases with the duration, meaning the longer the duration, the wider the window, resulting in more stable cardiopulmonary synchronization and a greater contribution to overall cardiopulmonary synchronization. This is calculated in the parameter space.n , m, d Below, the weighted cardiopulmonary synchronization curve is: (6).

[0019] Further, S204, specifically: for the synchronization ratio n:m and sensitivity threshold d For different values, the update rule for DSC is: if in the current parameter combination ( n h , m h , d i The weighted synchronization strength at a certain point is obtained. Greater than the previous parameter combination ( n k , m k , d j Synchronization intensity obtained under ) Then DSC ( t k Update to the result under the current parameter combination; otherwise, leave it unchanged. Iterate through all parameter spaces and find... t k The strongest evidence of synchronization between points.

[0020] Furthermore, S3, the specific steps are as follows: Since the dynamic synchronization curve DSC generated by S2 is based on the R peak time The DSC sequence is a non-uniform time series, while the emotion signal is a uniformly sampled continuous waveform with different sampling rates. To address this, both sequences are resampled to the same frequency. Interpolation is used to resample both the DSC sequence and the emotion annotation sequence to a frequency of 20 Hz, consistent with the emotion sampling rate, resulting in a time-synchronized standardized sequence. and , It includes two time series: arousal and valence.

[0021] Furthermore, S4, the specific steps are as follows: First, the joint probability distribution is estimated by counting the frequency of different pattern combinations in the statistical symbol sequence; let... and They are respectively t Given the symbol states of the source and target signals at given times, estimate the following probability density function: Target historical marginal probability: Joint probability of the target's own state: Source-target historical joint probability: Joint probability of the entire system: ; Based on the Shannon entropy calculated using the above probability density, the formula for the symbolic transfer entropy from source to target is: (11) in, Shannon entropy (based on the corresponding probability density) This formula precisely quantifies the additional information that the history of the source signal provides for predicting the future of the target by calculating the algebraic sum of the joint entropy.

[0022] Another embodiment of the present invention provides a quantitative system for the cardiopulmonary-emotion interaction mechanism based on symbolic transfer entropy, comprising the following modules: The acquisition module is used to collect the subjects' ECG and respiratory signals through wearable devices while the subjects watch emotionally evoked videos, and to record the changes in their subjective emotional valence and arousal using a continuous annotation tool based on a two-dimensional ring model, thereby obtaining an emotional annotation sequence. The dynamic DSC generation module combines ECG and respiratory signals to dynamically quantify changes in cardiopulmonary synchrony over a period of time and generate a dynamic cardiopulmonary synchrony curve (DSC). The DSC is based on the R peak time. Non-uniform time series; The synchronization module is used to resample the DSC and sentiment-labeled sequences to a frequency consistent with the sentiment sampling rate, thereby obtaining a time-synchronized standardized sequence. and ; STE calculation module, used for calculation and The directed information flow between them, namely the symbolic transfer entropy (STE) value; The significance test module uses the substitution data method to perform nonparametric statistical tests to establish the significance of causal connections. It establishes the causal prediction relationship between signals by comparing the bidirectional information flow characteristics, thereby determining the dominant driving direction and identifying the driving signal and response signal. The event-related analysis module identifies moments of significant change in the driving signal based on the determined dominant driving direction, defines abrupt events, and obtains the response waveforms of the corresponding driving events in the response signals. Based on the changing trend of the response waveforms relative to the driving events, it constructs state classification criteria to distinguish complex emotions with different inducing mechanisms.

[0023] Beneficial effects: Compared to traditional cardiopulmonary synchronization methods, the DSC index of this invention can pinpoint cardiopulmonary synchronization results to a single heartbeat cycle, capturing transient cardiopulmonary synchronization. Through multi-parameter spatial fusion of cardiopulmonary synchronization ratio and threshold, it can effectively distinguish between true phase locking and "pseudo-synchronization" caused by frequency proximity, exhibiting strong noise resistance. Furthermore, this invention provides a continuous coupling strength measure, rather than a binary (present / absent) classification, making it highly efficient, reliable, and easily software-based. Attached Figure Description

[0024] Figure 1 For technical flowcharts; Figure 2 Flowchart for dynamic cardiopulmonary synchronization curve (DSC) calculation; Figure 3 Synchronous cardiopulmonary results; Figure 4 The results of symbolic transfer entropy quantification under different emotions. Detailed Implementation

[0025] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0026] To quantify the interaction between cardiopulmonary synchronization and emotional experience, we propose a quantification method based on the cardiopulmonary-emotion interaction mechanism using symbolic transfer entropy. Figure 1 The technical flowchart of this invention is as follows: Step 1) Data Acquisition and Signal Preprocessing: First, while the subjects watched the emotion-evoking video, their electrocardiogram (ECG) and respiratory signals were acquired using wearable devices (Thought Technology SA9306 ECG sensor and Thought Technology SA9311M respiratory sensor). The changes in their subjective emotional valence and arousal were recorded in real time using a continuous annotation tool based on a two-dimensional circular model (such as a joystick, slider, or mouse tracking interface) to obtain the emotional signals.

[0027] For ECG signals, a 32nd-order FIR filter (cutoff frequency 3~45 Hz) was used to filter and remove baseline drift. Then, the R-peak time points were extracted using a QRS wave detection algorithm and processed into a point process sequence. For respiratory signals, baseline drift and high-frequency noise were first removed. A bandpass filter (0.05~0.5 Hz) was then used to process the signal and extract the frequency band containing the respiratory signal. Finally, the Hilbert Transform was used to calculate its analytic signal, thereby obtaining the continuous instantaneous respiratory phase.

[0028] Step 2) Calculation of Dynamic Cardiopulmonary Synchronization Curve (DSC) The purpose of this step is to dynamically quantify changes in cardiopulmonary synchrony levels over a period of time and generate a result curve, preparing for subsequent joint analysis with emotion. The technical flowchart for this part is shown below. Figure 2 As shown.

[0029] Automatic filtering of candidate synchronization ratios Considering individual differences and the non-stationarity of physiological states, this invention abandons the traditional practice of manually specifying synchronization ratios and adopts a dynamic expansion screening strategy based on frequency ratios. First, for each subject, the ratio of the average heart rate (average heartbeats per minute throughout the recording) to the average respiratory rate (average respiratory rates per minute throughout the recording) is calculated and rounded to the nearest integer to obtain the initial synchronization ratio. (For example, if the ratio of average heart rate to average respiratory rate is 3.27, then the initial synchronization ratio is...) n center : 1 = 3:1). Based on this, select the integer ratios that are adjacent to each other on the left and right ( and but The baseline candidate set for (i.e., 1 breath) includes ( , , m is the heart rate parameter in the synchronization ratio, and n is the respiratory parameter in the synchronization ratio.

[0030] To cover more complex physiological coupling patterns, m The range of values ​​for is extended to higher-order cases, i.e. m The values ​​range from 1 to 3. For each... m The value, its corresponding n The range of values ​​for is determined by the boundary expansion of the benchmark candidate set: the lower bound is . The upper boundary is Select the one that falls within [ , All integers within the numerical range k , construct the current m All candidate ratios under the value k : m For example, if the initial synchronization ratio is 3:1, then in At that time, the candidate ratios were 4:2, 5:2, 6:2, 7:2, and 8:2; When the candidate ratios are 6:3, 7:3, 8:3, 9:3, 10:3, 11:3, and 12:3, the advantage of this screening mechanism is that it constructs a search domain centered on the baseline cardiopulmonary synchronization ratio in the parameter space, which can automatically adapt to the baseline characteristics of the subjects and systematically cover higher-order synchronization ratios such as 5:2 and 7:2 that are common in complex physiological regulation, thereby ensuring the comprehensiveness and accuracy of the detection [4].

[0031] Constructing a multidimensional dynamic quantitative analysis framework The core of this invention lies in establishing a dynamic analysis framework that no longer relies on a single parameter but searches for evidence of phase locking in three dimensions. These three dimensions are: (1) cardiopulmonary synchronization ratio (Ratio), which iterates through all candidate ratios n:m selected in step 2; (2) sensitivity threshold (Threshold), which abandons the traditional comparison of a single fixed threshold with phase and sets a sensitivity threshold δ with a value range of [1,20] and a step size of 1; and (3) scale, which sets different observation window widths. L (i.e., the number of respiratory cycles included in a single synchronized radiograph assessment), such as L = 2,3,4,5, shorter scales are used to capture transient locks, and longer scales are used to confirm stable synchronization.

[0032] Multi-scale synchronous quantization To quantify the dynamic changes in cardiopulmonary synchronization, we employed a sliding window strategy to calculate synchronization intensity. Combining the ECG and respiratory signals obtained in step 1, we constructed a cardiopulmonary synchronization map, first calculating the time corresponding to each R peak. t k Relative phase sequence relative to the respiratory cycle F r : (1) in, f resp This is the respiratory phase. m This is the synchronization ratio parameter. This process converts two physiological signals from different modalities into a unified phase space, laying the foundation for subsequent synchronization analysis; relative phase sequence F r ( t k Defined in formula (1), the sequence length is N ,Right now k = 1, 2, …, N Regarding the time scale L (Values ​​are 2, 3, 4, 5), in the synchronization ratio n : mBelow, define a sliding window. W i = { F r ( t i ), F r ( t i+1 ), …, F r ( t i+n L-1 )}, i = 1, 2, …, N - n L +1, calculate the maximum deviation of the relative phase within each window, which reflects the stability of phase locking. The specific process is as follows: (1) Grouping data within a window (defining "rows"): In Under synchronization ratio, sliding window W i Contains A series of consecutive relative phase points, which should be distributed in n On different phase branches. Therefore, first, the data points within the window are arranged according to the index modulus. n Group it and split it into n Subgroups (i.e.) n (line), each line L 1 point (i.e., window width is ) L Each row of data set Defined as: , j = 0, 1, …, n -1,(2) in F r ( t k ) is the relative phase value within the window.

[0033] (2) Calculate the maximum phase difference within a "row": due to the relative phase F r ( t k ) is defined in For periodic variables on an interval, direct subtraction cannot reflect the true phase proximity (e.g., at the boundary of the period, the phase...). and (They are actually very close). Therefore, for any two points in time within a "row" ka and k b , k a , k b {1, 2, …, L},and k a ≠ k b Their relative phases are respectively F r ( t ka )and F r ( t kb Define the absolute difference between the two and the relationship between this difference and the period. m The smaller of the complements is the phase difference between the two points, that is: , k a , k b {1, 2, …, L},and k a ≠ k b (3) in, For Euclidean distance, To bridge the complementary distance across the periodic boundaries, this step eliminates the jump interference caused by phase periodicity. The phase difference between every two points within a "row" is calculated, and the maximum value is taken as the maximum phase difference of that "row" (the "horizontal" degree of the synchronization line), denoted as . : , k a , k b {1, 2, …, L}, k a ≠ k b (4) (3) Calculate the maximum value between "rows": After the maximum phase difference of each row in the sliding window is calculated, the maximum phase difference of the window is defined as the maximum phase difference of the "row". , j = 0, 1, …, n The maximum value of -1: (5) in, m , n Synchronization ratio parameter L The scale (i.e., window width) is used for each sliding window. W i Inside n "Line", each "line" has L Each element (relative phase).

[0034] (4) Determining the synchronization state: Represents the current window W i The worst-case phase-locked condition is defined if and only if the value is less than the sensitivity threshold. Only when the tolerance limit is determined will a judgment be made. W i In a synchronized state. This invention introduces an iterative update mechanism, abandoning the traditional selection of a single fixed threshold, and sets a sensitivity threshold δ, with a value from 1 to 20 and a step size of 1. Define the sequence. S L In scale L The following is a synchronized time series with a length of N For each sliding window W i If its maximum phase difference D( W i , L , n , m The tolerance is less than the current sensitivity threshold. m / ( n · d If the window is not found, then the decision window will be determined. W i At the threshold level d If we select "Synchronize", then we can synchronize the time series. S L Zhongyu W i The corresponding time point S L ( t i ), S L ( t i+1 ), …, S L ( t i+n L-1 Then assign the value d If the maximum phase difference D within the window ( W i , L ,n , m Not less than the tolerance corresponding to the current threshold m / ( n · d ), then S L ( t i ), S L ( t i+1 ), …, S L ( t i+n L-1 The value is assigned to 0. Synchronize the time series. S L Refresh mechanism: When the sliding window is refreshed by... W i Become W i+1 When, if D( W i+1 , L , n , m Less than m / ( n · d If ), then record S L ( t i+1 ), S L ( t i+2 ), …, S L ( t i+n L )for d If D( W i+1 , L , n , m Not less than m / ( n · d If ), then retain. S L ( t i+1 ), …, S L ( t i+n L-1 The value of ) remains unchanged, and the record is maintained. SL ( t i+n L The value is 0; that is, in the synchronized time series S L The maximum synchronization threshold is retained during the sliding process. When taking... i For 1, 2, …, N - n L When +1 is added, a synchronization time series reflecting the synchronization level can be obtained. S L .

[0035] (5) Multi-scale fusion: different time scales L = 2, 3, 4, 5, which correspond to different synchronization time series. This indicates different cardiopulmonary synchronization durations; the longer the duration (i.e., the wider the window), the better. L The larger the window width, the more stable the phase locking of the cardiopulmonary signal can remain within a longer time window, i.e., a more stringent level of cardiopulmonary synchronization; L The smaller the value, the better it can capture transient changes in cardiopulmonary synchronization. To achieve a balance between transient changes and stable synchronization, we... S L The sequences are weighted, and a weight vector is defined. Q = [ oh L1 , oh L2 , …, oh Ll ] l For scale L The number of possible values ​​(four values ​​in total), and S L correspond, oh Li The value of varies with i The duration increases with the window width, meaning the longer the duration (the wider the window), the more stable the cardiopulmonary synchrony and the greater its contribution to overall cardiopulmonary synchrony. For example, if... Q = [0.05, 0.2, 0.25, 0.5]. The calculated values ​​in the parameter space are... n , m, d Below, the weighted cardiopulmonary synchronization curve is: (6) Iterative update of synchronization curve (DSC) For the synchronization ratio n:m and sensitivity threshold dFor different values ​​of , iterating through the above algorithm, the update rule for DSC is: if in the current parameter combination ( n h , m h , d i The weighted synchronization strength at a certain point is obtained. Greater than the previous parameter combination ( n k , m k , d j Synchronization intensity obtained under ) Then DSC ( t k Update to the result under the current parameter combination; otherwise, leave it unchanged. Traversing the entire parameter space allows you to find... t k The strongest evidence of synchronization of points (i.e., all points k=1-N) t k The strongest synchronization evidence at a point constitutes the dynamic synchronization curve (DSC), i.e., the maximum sensitivity threshold. The complete algorithm flow is as follows: Figure 1 As shown. This process not only solves the synchronization ratio competition problem, but also improves the temporal resolution of cardiopulmonary synchronization, theoretically accurate to a single heartbeat cycle. For example... Figure 3 As shown, the obtained DSC is a continuously changing time series that can intuitively represent the strength and stability of cardiopulmonary phase coupling, effectively distinguishing between true phase locking and "pseudo-synchronization" caused by accidental frequency proximity.

[0036] Step 3) Sequence Symbolization and Phase Space Reconstruction Since the Dynamic Synchronization Curve (DSC) generated in Step 2 is based on the R peak time ( The DSC sequence is a non-uniform time series, while the emotion signal is a uniformly sampled continuous waveform with different sampling rates. To calculate the transfer entropy, both need to be resampled to the same frequency. In this invention, we use interpolation to resample both the DSC sequence and the emotion sequence (Valence / Arousal) to a frequency (20 Hz) consistent with the emotion sampling rate, resulting in a time-synchronized standardized sequence. and (Includes two time series: arousal and valence).

[0037] To calculate the Symbolic Transfer Entropy (STE), it is first necessary to transform the continuous time series into a discrete symbolic sequence and reconstruct its dynamic state space. This invention adopts a "symbolization first, reconstruction later" strategy.

[0038] Set the local embedding dimension of the symbolization process to , and the embedding delay to . In this embodiment, in order to capture local dynamic characteristics while avoiding an overly large symbol space, take , .

[0039] For each moment t in the time series x(t), construct a local vector of length : . Compare the relative magnitude relationships of the internal elements of the vector and map it to a unique symbol , with the rule: map according to the permutation of the indices sorted by their magnitudes. For , there are 3! = 6 kinds of rising and falling trend patterns (such as "monotonically increasing", "monotonically decreasing", "rising first and then falling", etc.). For the possible equal-value situations that may occur in cardiopulmonary synchronization or emotional signals (e.g., x(t) = x(t+1)<x(t+2)), the present invention does not introduce random noise, but defines these specific equal-value structures as independent patterns. For , there are a total of 13 possible rank patterns including equal-value situations. Assign a unique integer index (such as 1~13) to the identified patterns. Thus, the original continuous sequence is converted into a discrete symbol sequence and , where each symbol represents the local waveform structure at that moment.

[0040] To calculate transfer entropy, it is necessary to determine how much historical information in the symbol sequence is required to effectively predict future states. Therefore, it is necessary to perform phase space reconstruction on the generated symbol sequences and to determine the optimal reconstruction delay ( ) and reconstruction dimension ( d ).

[0041] We use the average mutual information method (AMI) to determine the reconstruction delay , and the purpose of using this method is to determine the optimal time interval between two successive sampling points in the symbol sequence to eliminate information redundancy. The specific process is as follows: For the discrete symbol sequence , calculate the average mutual information between it and the delayed sequence : (7) T is the embedding delay, where is 's probability density, is the joint probability density, estimated by the histogram method. Calculate a series of differentT Value Search The first local minimum. The corresponding local minimum. T That is, the optimal embedding delay This choice ensures that the components of the reconstructed vector are neither overly correlated (information redundancy) nor completely uncorrelated (structure loss). Next, the False Nearest Neighbors (FNN) method is used to determine the embedding dimension. d This is because when the reconstruction dimension is too low, trajectory points that are originally far apart in the complete phase space may appear very close due to projection overlap, forming "false neighbor points." The goal of the FNN method is to find a minimum dimension to eliminate these false neighbors caused by projection. The specific process is as follows: Let's assume... d In 3D space, a point The nearest neighbor is Calculate them in Distance variation in 2D space. Defining distance ratios. R i : (8) The denominator is the Euclidean distance in d-dimensional space, and the numerator is the distance increment on the newly added coordinate axis after adding one dimension. It is a point Projection in d+1 dimensional space The nearest neighbor is Set threshold R tol (Usually 10-15, but 10 is used in this invention). If R i ( d )> R tol If the neighboring point is not found to be a false neighbor, then the neighboring point is determined to be a false neighbor. Traverse from smallest to largest. d (like d = 1, 2, …, 10), calculate the percentage of false neighbors in each dimension. Select the point where the percentage of false neighbors first drops close to zero (below a set threshold, such as 5%). d The value is used as the optimal embedding dimension.

[0042] Based on the determined reconfiguration delay and reconstructed dimension d This invention introduces a source-target action delay parameter to construct a historical state vector for symbolic transfer entropy (STE) calculation. u It supports using independent embedding parameters for source and target sequences to accurately capture causal effects under different physiological-psychological lags. The history vector of the source signal is: (9) The historical vector of the target signal is: (10) It is the reconstruction delay corresponding to the source signal. It is the reconstructed dimension corresponding to the source signal; It is the reconstruction delay corresponding to the source signal. It is the reconstructed dimension corresponding to the source signal.

[0043] Step 4) Symbolic Transfer Entropy (STE) Calculation: Using the historical vector obtained in Step 3, the directed information flow between cardiopulmonary synchrony (DSC) and emotional state is calculated, i.e., the symbolic transfer entropy. First, the joint probability distribution is estimated by statistically analyzing the frequency of different pattern combinations in the symbol sequence. Let... and They are respectively t Given the symbol states of the source and target signals at given times, estimate the following probability density function: Target historical marginal probability: Joint probability of the target's own state (historical + future): Joint probability of source-target history (target history + source history): Joint probability of the entire system (target history + future + source history): Based on the Shannon entropy calculated using the above probability density, the formula for the symbolic transfer entropy from source to target is: (11) in, Shannon entropy (based on the corresponding probability density) This formula precisely quantifies the amount of additional information that the history of the source signal provides for predicting the future of the target by calculating the algebraic sum of the joint entropy.

[0044] Step 5) Statistical Significance Test and Directionality Determination: Verifying the significance of a specific direction is to confirm that the past state of the source signal can provide additional, directional information gain for predicting the future state of the target signal. Since the STE value may have non-zero bias when the sample size is limited, this invention uses the Surrogate Data Method to perform nonparametric statistical tests to establish the significance of the causal connection. The specific process is as follows: Maintaining... Unchanged, will The time sequence is randomly shuffled to generate a set of alternative sequences. Y surr This operation broke and The original temporal correlation between them was preserved. The statistical distribution characteristics. Repeat the above process (>100 times, 200 times in this invention), calculating the substitution sequence for each iteration. X surr right The transition entropy value is used to construct the null distribution of the STE. Then, significance is determined by setting the significance level (e.g., ...). If the original sequence is calculated as follows: Values ​​greater than 95% of the zero distribution th Percentiles indicate that the information flow in that direction is statistically significant.

[0045] This invention relates to and The STE calculation and null distribution significance test described above are performed in both opposite directions. The causal predictive relationship between the signals is established by comparing the bidirectional information flow characteristics: if... It is statistically significant, and its value (information transmission volume) is significantly greater than that in the opposite direction. (Or, if the opposite direction fails the significance test), it indicates that the information flow exhibits strong asymmetry. This is used to determine the driving signal and the response signal.

[0046] Step 6) Analysis of the operational mode based on event correlation analysis method. This step utilizes the source-target delay parameters obtained from previous calculations. u By capturing the "drive-response" waveform characteristics at a microscopic time scale, this study further analyzes the specific interaction patterns between the cardiopulmonary system and the emotional system. First, based on the dominant drive direction determined in step 5, the study identifies moments of significant change in the drive signal (such as the cardiopulmonary synchronized DSC curve). Based on this, we can define various abrupt events, such as defining an "enhancement event" as a segment where the signal amplitude rapidly increases, and defining an "inhibition event" as a segment where the signal amplitude rapidly decreases (the specific definition depends on the signal itself; for example, in this invention, we define a positive first-order difference in the DSC curve lasting 5 seconds as a "rapid increase" state, and a negative first-order difference in the DSC curve lasting 5 seconds as a "rapid decrease" state). To accurately capture the specific impact of the drive event on the response signal, the optimal source-target interaction delay obtained in step 5 is utilized. u Perform timeline correction and construct an event-related observation window. For each identified event... t The driving event at a given moment, within the corresponding time lag window of the response signal (such as an emotional sequence). Data segments are extracted and averaged to obtain typical response waveforms under the abrupt change event. Based on the trend of the response waveform relative to the driving event (positive or negative correlation), a state classification criterion is constructed to distinguish complex emotions with different inducing mechanisms. For example, in a scenario where the driving signal is a DSC sequence, a high arousal response triggered by an "ascending event" is representative of the emotional state of "fear." Conversely, a high arousal response triggered by a "descending event" often reflects a "pleasure" or "relaxation" state dominated by respiratory sinus arrhythmia (RSA). Therefore, by combining the type of driving event (ascending / descending) with the specific trajectory of the response waveform, a classification feature space with high physiological interpretability can be established, effectively overcoming the performance bottleneck of traditional emotion computing in recognizing complex emotions.

[0047] Figure 4 This paper presents statistical results obtained by processing data from four representative emotional states (interesting, bored, relaxed, and fearful) in the Circumplex Model of Emotion using the method described in this invention. These four states cover the four typical quadrants of the emotional space: interesting (high arousal / positive valence), bored (low arousal / negative valence), relaxed (low arousal / positive valence), and fearful (high arousal / negative valence). The figure visually illustrates the strength of the bidirectional causal connection between cardiopulmonary synchrony and emotion. In all emotional states, the red node (cardiopulmonary synchrony → emotion) is located to the right of the cyan node (emotion → cardiopulmonary synchrony), and there is a significant distance between them. This indicates that this method successfully captures and quantifies the asymmetry in the cardiopulmonary-emotion interaction. Furthermore, the STE distribution differs under different emotions. For example, in the "interesting" state, the value of emotion → cardiopulmonary synchrony (cyan) is significantly higher than in the "bored" state. This demonstrates that this method can not only determine the dominant direction but also sensitively detect the enhanced central regulatory mechanisms under specific emotions (such as high arousal and positive valence emotions).

[0048] This technology innovatively quantifies the information flow between cardiopulmonary synchronization and emotional experience. Combining the high temporal resolution of DSC with the robustness of STE to noise and nonlinearity, it is suitable for analyzing complex physiological-psychological signals and is easy to software-encode.

[0049] An embodiment of the present invention also provides a system for performing a quantification method for the cardiopulmonary-emotion interaction mechanism based on symbolic transfer entropy, specifically including the following modules: The acquisition module is used to collect the subjects' ECG and respiratory signals through wearable devices while the subjects watch emotionally evoked videos, and to record the changes in their subjective emotional valence and arousal using a continuous annotation tool based on a two-dimensional ring model, thereby obtaining an emotional annotation sequence. The dynamic DSC generation module combines ECG and respiratory signals to dynamically quantify changes in cardiopulmonary synchrony over a period of time and generate a dynamic cardiopulmonary synchrony curve (DSC). The DSC is based on the R peak time. Non-uniform time series; The synchronization module is used to resample the DSC and sentiment-labeled sequences to a frequency consistent with the sentiment sampling rate, thereby obtaining a time-synchronized standardized sequence. and ; STE calculation module, used for calculation and The directed information flow between them, namely the symbolic transfer entropy (STE) value; The significance test module uses the substitution data method to perform nonparametric statistical tests to establish the significance of causal connections. It establishes the causal prediction relationship between signals by comparing the bidirectional information flow characteristics, thereby determining the dominant driving direction and identifying the driving signal and response signal. The event-related analysis module identifies moments of significant change in the driving signal based on the determined dominant driving direction, defines abrupt events, and obtains the response waveforms of the corresponding driving events in the response signals. Based on the changing trend of the response waveforms relative to the driving events, it constructs state classification criteria to distinguish complex emotions with different inducing mechanisms.

[0050] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A quantitative method for the cardiopulmonary-emotion interaction mechanism based on symbolic transfer entropy, characterized in that, Includes the following steps: S1: While watching the emotionally evoked video, the subjects' ECG and respiratory signals were collected through wearable devices, and their subjective emotional valence and arousal changes were recorded using a continuous annotation tool based on a two-dimensional ring model to obtain an emotional annotation sequence. S2: Combining ECG and respiratory signals, dynamically quantify the changes in cardiopulmonary synchronization over a period of time and generate a result curve, producing a dynamic cardiopulmonary synchronization curve (DSC). The DSC is based on the R peak time. Non-uniform time series; S3: Resample the DSC and sentiment-labeled sequences to a frequency consistent with the sentiment sampling rate to obtain a time-synchronized standardized sequence. and ; S4: Calculation and The directed information flow between them, namely the symbolic transfer entropy (STE) value; S5: Use alternative data method to perform nonparametric statistical tests to establish the significance of causal connections. By comparing the bidirectional information flow characteristics, establish the causal prediction relationship between signals, thereby determining the dominant driving direction and identifying the driving signal and response signal. S6: Based on the determined dominant driving direction, identify the moment when a significant change occurs in the driving signal to define a mutation event, and obtain the response waveform of the corresponding driving event in the response signal. Based on the changing trend of the response waveform relative to the driving event, construct a state classification criterion to distinguish complex emotions with different inducing mechanisms.

2. The method for quantifying the cardiopulmonary-emotion interaction mechanism based on symbolic transfer entropy as described in claim 1, characterized in that, While the subjects watched the emotionally evoked videos, their electrocardiogram (ECG) and respiratory signals were collected using wearable devices, and then processed as follows: S1-1: For ECG signals, a 32nd-order FIR filter is used to filter and remove baseline drift. Then, the R-peak time points are extracted using the QRS wave detection algorithm and processed into a point process sequence. S1-2: For the respiratory signal, firstly remove baseline drift and high-frequency noise, then use a bandpass filter to process the signal and extract the frequency band where the respiratory signal is located, and then use Hilbert transform to calculate its analytic signal, thereby obtaining the continuous instantaneous respiratory phase.

3. The method for quantifying the cardiopulmonary-emotion interaction mechanism based on symbolic transfer entropy as described in claim 2, characterized in that, The process of combining ECG and respiratory signals to dynamically quantify changes in cardiopulmonary synchronization over a period of time and generate a dynamic cardiopulmonary synchronization curve (DSC) involves the following steps: S201: Screening candidate synchronization ratio; S202: Construct a multidimensional dynamic quantitative analysis framework, including three dimensions: (1) cardiopulmonary synchronization ratio (Ratio), which iterates through all candidate synchronization ratios; (2) sensitivity threshold, which sets a sensitivity threshold δ with a value range of [1, 20] and a step size of 1; and (3) scale, which sets different observation window widths. L ; S203: Combine ECG and respiratory signals to construct a cardiopulmonary synchronization map and perform multi-scale synchronization quantization to obtain a cardiopulmonary synchronization curve; S204: For different values ​​of cardiopulmonary synchronization ratio and sensitivity threshold, the above algorithm is traversed to iteratively update the cardiopulmonary synchronization curve and obtain the dynamic cardiopulmonary synchronization curve DSC. The obtained dynamic cardiopulmonary synchronization curve DSC is a continuously changing time series, which is used to intuitively represent the strength and stability of cardiopulmonary phase coupling.

4. The method for quantifying the cardiopulmonary-emotion interaction mechanism based on symbolic transfer entropy as described in claim 3, characterized in that, S201: Screening candidate synchronization ratios, the specific steps are as follows: First, for each subject, the ratio of the mean heart rate to the mean respiratory rate throughout the entire recording was calculated, and then rounded to the nearest integer to obtain the initial synchronization ratio. Based on this, select the integer ratios of its left and right adjacent values ​​( and but That is, the baseline candidate set for one breath includes ( , , ; m is the heart rate parameter in the synchronization ratio, and n is the respiratory parameter in the synchronization ratio; m The value of is iterated from 1 to 3, for each . m The value, its corresponding n The range of values ​​for is determined by the boundary expansion of the benchmark candidate set: the lower bound is . The upper boundary is ; Select the one that falls on the [ , All integers within the numerical range k , construct the current m All candidate synchronization ratios under the value k : m .

5. The method for quantifying the cardiopulmonary-emotion interaction mechanism based on symbolic transfer entropy as described in claim 4, characterized in that, The specific steps of S203 are as follows: Combining the electrocardiogram and respiratory signals obtained in step 1, a cardiopulmonary synchronization map is constructed. First, the time corresponding to each R peak is calculated. t k Relative phase sequence relative to the respiratory cycle Φ r : （1） in, φ resp This is the respiratory phase. m This is the synchronization ratio parameter; this process converts two physiological signals from different modalities into a unified phase space, laying the foundation for subsequent synchronization analysis. relative phase sequence Φ r ( t k Defined in formula (1), the sequence length is N ,Right now k = 1, 2, …, N Regarding the time scale L (Values ​​are 2, 3, 4, 5), in the synchronization ratio n : m Below, define a sliding window. W i = { Φ r ( t i ), Φ r ( t i+1 ), …, Φ r ( t i+n L-1 )}, i = 1, 2, …, N - n L +1, calculate the maximum deviation of the relative phase within each window. This deviation reflects the stability of phase locking. The specific process is as follows: (1) Grouping data within a window, defining "rows": In Under synchronization ratio, sliding window W i Contains A series of consecutive relative phase points, which should be distributed in n On different phase branches, first, the data points within the window are sorted by index modulus. n Group it and split it into n Subgroups, i.e. n Lines, each line L The number of points represents the window width. L Each row of data set Defined as: ， j = 0, 1, …, n -1（2） in Φ r ( t k () represents the relative phase value within the window; (2) Calculate the maximum phase difference within a "row": For any two time points within a "row" k a and k b , k a , k b {1, 2, …, L },and k a ≠ k b Their relative phases are respectively Φ r ( t ka )and Φ r ( t kb Define the absolute difference between the two and the relationship between this difference and the period. m The smaller of the complements is the phase difference between the two points, that is: （3） in, k a , k b {1, 2, …, L },and k a ≠ k b , For Euclidean distance, To eliminate the jump interference caused by phase periodicity, the complementary distance across the periodic boundary is calculated. The phase difference between every two points within a "row" is then taken as the maximum phase difference for that "row," denoted as . : ， k a , k b {1, 2, …, L}， k a ≠ k b （4） (3) Calculate the maximum value between "rows": After the maximum phase difference of each row in the sliding window is calculated, the maximum phase difference of the window is defined as the maximum phase difference of the "row". , j = 0, 1, …, n The maximum value of -1: （5） in, m , n Synchronization ratio parameter L The dimension is the width of the window, and each sliding window... W i Inside n "Lines", each "line" has L One element; (4) Determining the synchronization state: Represents the current window W i The worst-case phase-locked condition is defined if and only if the value is less than the sensitivity threshold. Only when the tolerance limit is determined will a judgment be made. W i In a synchronized state, an iterative update mechanism is introduced, abandoning the traditional single fixed threshold selection, setting a sensitivity threshold δ with a value from 1 to 20 and a step size of 1; defining a sequence. S L In scale L The following is a synchronized time series with a length of N For each sliding window W i If its maximum phase difference D( W i , L , n , m The tolerance is less than the current sensitivity threshold. m / ( n · δ If the window is not found, then the decision window will be determined. W i At the threshold level δ Next, "synchronize", then synchronize the time series. S L Zhongyu W i The corresponding time point S L ( t i ), S L ( t i+1 ), …, S L ( t i+n L-1 Then assign the value δ If the maximum phase difference D within the window ( W i , L , n , m Not less than the tolerance corresponding to the current threshold m / ( n · δ ), then S L ( t i ), S L ( t i+1 ), …, S L ( t i+n L-1 The value is assigned to 0; Synchronous Time Series S L Refresh mechanism: When the sliding window is refreshed by... W i Become W i+1 When, if D( W i+1 , L , n , m Less than m / ( n · δ If ), then record S L ( t i+1 ), S L ( t i+2 ), …, S L ( t i+n L )for δ If D( W i+1 , L , n , m Not less than m / ( n · δ If ), then retain. S L ( t i+1 ), …, S L ( t i+n L-1 The value of ) remains unchanged, and the record is maintained. S L ( t i+n L The value is 0; that is, in the synchronized time series S L The maximum synchronization threshold is retained during the sliding process; when taking i For 1, 2, …, N - n L When +1 is added, a synchronization time series reflecting the synchronization level can be obtained. S L ; (5) Multi-scale fusion: different time scales L ,right S L The sequences are weighted, and a weight vector is defined. Q = [ ω L1 , ω L2 , …, ω Ll ] l For scale L The number of possible values, and S L correspond, ω Li The value of varies with i The window width increases with the duration, meaning the longer the duration, the wider the window, resulting in more stable cardiopulmonary synchronization and a greater contribution to overall cardiopulmonary synchronization. This is calculated in the parameter space. n , m, δ Below, the weighted cardiopulmonary synchronization curve is: (6)。 6. The method for quantifying the cardiopulmonary-emotion interaction mechanism based on symbolic transfer entropy as described in claim 5, characterized in that, Specifically, S204 refers to: for the synchronization ratio n:m and the sensitivity threshold... δ For different values, the update rule for DSC is: if in the current parameter combination ( n h , m h , δ i The weighted synchronization strength at a certain point is obtained. Greater than the previous parameter combination ( n k , m k , δ j Synchronization intensity obtained under ) Then DSC ( t k Update to the result under the current parameter combination; otherwise, leave it unchanged. Iterate through all parameter spaces and find... t k The strongest evidence of synchronization between points.

7. The method for quantifying the cardiopulmonary-emotion interaction mechanism based on symbolic transfer entropy as described in claim 6, characterized in that, The specific steps of S3 are as follows: Since the dynamic synchronization curve DSC generated by S2 is based on the R peak time The DSC sequence is a non-uniform time series, while the emotion signal is a uniformly sampled continuous waveform with different sampling rates. To address this, both sequences are resampled to the same frequency. Interpolation is used to resample both the DSC sequence and the emotion annotation sequence to a frequency of 20 Hz, consistent with the emotion sampling rate, resulting in a time-synchronized standardized sequence. and , It includes two time series: arousal and valence.

8. The method for quantifying the cardiopulmonary-emotion interaction mechanism based on symbolic transfer entropy as described in claim 7, characterized in that, The specific steps of S4 are as follows: First, the joint probability distribution is estimated by counting the frequency of different pattern combinations in the statistical symbol sequence; let... and They are respectively t Given the symbol states of the source and target signals at given times, estimate the following probability density function: Target historical marginal probability: Joint probability of the target's own state: ; Source-target historical joint probability: Joint probability of the entire system: ; Based on the Shannon entropy calculated using the above probability density, the formula for the symbolic transfer entropy from source to target is: (11) in, Shannon entropy (based on the corresponding probability density) This formula precisely quantifies the additional information that the history of the source signal provides for predicting the future of the target by calculating the algebraic sum of the joint entropy.

9. A quantitative system for cardiopulmonary-emotion interaction mechanism based on symbolic transfer entropy, characterized in that, Includes the following modules: The acquisition module is used to collect the subjects' ECG and respiratory signals through wearable devices while the subjects watch emotionally evoked videos, and to record the changes in their subjective emotional valence and arousal using a continuous annotation tool based on a two-dimensional ring model to obtain an emotional annotation sequence. The dynamic DSC generation module combines ECG and respiratory signals to dynamically quantify changes in cardiopulmonary synchrony over a period of time and generate a dynamic cardiopulmonary synchrony curve (DSC). The DSC is based on the R peak time. Non-uniform time series; The synchronization module is used to resample the DSC and sentiment-labeled sequences to a frequency consistent with the sentiment sampling rate, thereby obtaining a time-synchronized standardized sequence. and ; STE calculation module, used for calculation and The directed information flow between them, namely the symbolic transfer entropy (STE) value; The significance test module uses the substitution data method to perform nonparametric statistical tests to establish the significance of causal connections. It establishes the causal prediction relationship between signals by comparing the bidirectional information flow characteristics, thereby determining the dominant driving direction and identifying the driving signal and response signal. The event-related analysis module identifies abrupt events by recognizing moments of significant change in the driving signal based on the determined dominant driving direction. It also acquires the response waveforms of the corresponding driving events from the response signals. Based on the changing trends of the response waveforms relative to the driving events, it constructs state classification criteria to distinguish complex emotions with different inducing mechanisms.

Images